Existing targeted cystic fibrosis screening assays miss important pathogenic CFTR variants in the ethnically diverse US population.
To evaluate the analytic performance of a multiplex polymerase chain reaction (PCR)/capillary electrophoresis (CE) CFTR assay panel that simultaneously interrogates primary pathogenic variants of different ethnic/ancestral groups.
Performance characteristic assessment and variant coverage comparison of the panel with a focus on ethnicity-specific CFTR variants were performed. Sample DNA was primarily from whole blood or cell lines. Detection of CFTR carriers was compared across several commercially available CFTR kits and recommended variant sets based on panel content.
The panel interrogated 65 pathogenic CFTR variants representing 92% coverage from a recent genomic sequencing survey of the US population, including 4 variants with top 5 frequency in African or Asian populations not reflected in other targeted panels. In simulation studies, the panel represented 95% of carriers across the global population, resulting in a 6.9% to 19.0% higher carrier detection rate compared with 10 targeted panels or variant sets. Precision and sensitivity/specificity were 100% concordant. Multisite sample-level genotyping accuracy was 99.2%. Across PCR and CE instruments, sample-level genotyping accuracy was 97.1%, with greater than 99% agreement for all variant-level metrics.
The CFTR assay achieves 92% or higher coverage of CFTR variants in diverse populations and provides improved pan-ethnic coverage of minority subgroups of the US populace. The assay can be completed within 5 hours from DNA sample to genotype, and performance data exceed acceptance criteria for analytic metrics. This assay panel content may help address gaps in ancestry-specific CFTR genotypes while providing a streamlined procedure with rapidly generated results.
Cystic fibrosis (CF) is a potentially lethal autosomal recessive disease characterized by severe impairment of the respiratory system and digestive tract.1,2 CF occurs in approximately 1:3500 births in the United States, and at a higher rate in some northern European countries, with a lesser incidence in African and Asian populations.3,4 Variants in the CF transmembrane conductance regulator gene (CFTR) result in diverse aberrations in the protein responsible for chlorine and bicarbonate transport across epithelial cell membranes.4 Dysfunctional or even absent epithelial anion transporter protein results in variable CF disease phenotypes, including thicker mucus lining the bronchi, impaired pancreatic secretory function, and compromised intestinal absorption. Efforts to develop therapeutic strategies for CF depend on a comprehensive understanding of its underlying molecular mechanisms.1,2
Of the more than 2000 CFTR variants identified to date, 401 have been shown to cause CF as of 2022.5 Pathogenic CFTR variants occur at different allele frequencies in different populations.6 Carrier frequency within the US populace for CF-associated CFTR variants ranges from ≈1:30 in non-Hispanic Whites to ≈1:90 in Asians.3 In 2001 the American College of Medical Genetics (ACMG) CF Carrier Screening Working Group and the American College of Obstetricians and Gynecologists (ACOG) published Practice Guidelines that recommended genetic testing against a core panel of 25 variants for women in the United States who are pregnant or considering pregnancy.7,8 This list was revised to a minimum panel of 23 pathogenic variants (CF23) with prevalence in the United States reflecting greater than 0.1% of CF cases,9 and was reaffirmed by ACMG as recently as 2020.10
Reliance on the minimum CF23 panel risks missing CFTR variants that are increasingly common as a population diversifies in ethnic makeup.11 Indeed, a 2019 study used next-generation sequencing (NGS) to survey more than 115 000 individuals across the United States and identified 213 unique pathogenic CFTR variants, noting that CF23 screening alone would have missed 31% of at-risk pregnancies.12 Another NGS survey of more than 380 000 individuals in the United States detected 481 known or suspected pathogenic CFTR variants and showed marked differences in the constitution of the top 5 pathogenic CFTR variants across different ethnic groups, including many that would be missed by a CF23-only panel (Supplemental Table 1, see supplemental digital content at https://meridian.allenpress.com/aplm in the October 2024 table of contents).13 These observations underscore professional guideline updates that recommend expanding CFTR screening beyond the prototypical CF23 panel to capture variants associated with the unique ethnic makeup of the intended testing population.10,14,15 In 2020, a critical resource, the Genome Aggregation Database (gnomAD), identified high-confidence loss-of-function variants from 141 456 humans (aggregated whole-exome and whole-genome sequencing data).16 Because ancestry in these individuals was genetically determined instead of being self-reported, we hypothesized that this database would also be a valuable tool for comprehensively assessing pathogenic CFTR variant frequencies in heterogeneous populations. Indeed, in their most recent 2023 CF carrier screening recommendations, the ACMG leveraged gnomAD variant frequencies and calculated ancestry to update the minimum recommended set of variants to 100, with the objective of representing 95% of carriers in each of 6 ancestral populations.11
Here, we describe the panel design, workflow, and prospective performance evaluation of a targeted polymerase chain reaction (PCR)/capillary electrophoresis (CE) assay that simultaneously interrogates samples for 65 unique CFTR variants, including those in the CF23 panel. This 2-tube assay recognizes 63 unique pathogenic or likely pathogenic CFTR variants representing 92% coverage of the cumulative mutant allele frequency (MAF) in the United States.12 With the addition of phased PolyT/TG sizing, the full panel includes at least 1 pathogenic CFTR variant in more than 99% of current US patients with CF. We evaluated the analytic performance of the assay across different thermal cycler and CE configurations using standard metrics of accuracy, precision, specificity, and sensitivity. The variant panel was also compared to 10 other targeted multiplex CFTR assay panels for cumulative MAF coverage across multiple reference CFTR variant data sets and ancestral populations.
MATERIALS AND METHODS
Study Design and Ethical Oversight
This study had 2 objectives: (1) determine the representation of pathogenic variants in the US population from the listed panel content of targeted CFTR panels using multiple diverse (ethnicity and ancestry) prevalence data sets; and (2) assess the analytic performance of the 65-variant PCR/CE kit reagents (AmplideX PCR/CE CFTR kit, catalog Nos. A00076 and A00077 [Research Use Only], Asuragen Inc, Austin, Texas) following the manufacturer’s instructions for use. Prospective performance studies included sample- and variant-level accuracy, sensitivity, specificity, precision, and multisite evaluation. Metrics included positive, negative, and overall percent agreement (PPA, NPA, and OPA, respectively); positive predictive value (PPV); accuracy of zygosity determination for single-nucleotide variants (SNVs) and insertions/deletions (indels); and accuracy of PolyT/TG repeat length and phasing. Sample-level accuracy was determined by evaluating the assay genotyping accuracy compared to reference methods, with possible genotypes including wild-type (WT; 0 pathogenic CFTR variants), heterozygous (1 variant; carrier), compound heterozygous (2 variants; likely affected), homozygous (same variant on both alleles; affected), or multiple (3+ variants; affected). Additionally, the assay was tested across multiple thermal cycler and CE instrument configurations, across a range of input DNA quantities, and for stability of kit reagents and allele-specific PCR products during nominal workflow processing.
Assay Workflow
The AmplideX PCR/CE CFTR kit includes reagents for targeted in vitro multiplex nucleic acid amplification that simultaneously detects and identifies a panel of 65 CFTR variants (the target selection logic, process, and resources are detailed in the Supplemental Materials text and Supplemental Table 2). The assay is used to perform PCR-based genotype analysis on genomic DNA (20–60 ng/sample input; Supplemental Table 3) extracted from peripheral venous blood samples, and it is compatible with standard laboratory-validated PCR thermal cyclers, with amplicon resolution conducted on Applied Biosystem CE platforms. According to manufacturer instructions and this study, DNA sample-to-answer can be completed in ≈5 hours, with hands-on operator time comprising less than 1 hour.
This CFTR assay can genotype and identify the zygosity of all of the CF23 variants plus 11 additional CFTR variants within a single PCR tube (CFTR Primer Mix A) that together represent 86% of the cumulative MAF in the United States. A second PCR tube is used to detect 31 lower-frequency variants (CFTR Primer Mix B). Published MAFs for 63 of 65 CFTR variants (not including PolyT/TG) are provided in Table 1.
PCR and CE Instruments
A variety of thermal cyclers (Veriti, GeneAmp 9700, and ProFlex, all from ThermoFisher, Waltham, Massachusetts; and the C1000 Touch, Bio-Rad Laboratories Inc, Hercules, California) were used to amplify PCR products, which were subsequently resolved on various CE platforms (3130xl, 3500xL, 3730xl, and SeqStudio Genetic Analyzers; all from ThermoFisher) following manufacturer protocols for each instrument configuration.
CFTR PCR/CE Assay Data Processing
Results from the assay were analyzed using the AmplideX PCR/CE CFTR Analysis Module v1.1.0 within the AmplideX PCR/CE Reporter v3.0.3 software (Asuragen). The module automatically detects and reports amplicon peak size (bp) based on a ROX (6-carboxyl-X-rhodamine)-labeled size ladder, peak intensity (relative fluorescence units [RFUs]), and dye channel. The module associates each peak to a variant bin and reports zygosity for each variant, and it identifies sample-level information, such as genotype and total variants. The software detects peak position and dye channel of the PolyT/TG region to determine both repeat size and phase. The software processes and analyzes data directly from raw .FSA files produced by CE instruments, and it performs quality control (QC) checks for signal intensity and other common indicators of quality results. Results are displayed in the software in less than 5 minutes per batch.
Data were excluded from analysis if they failed any QC check not remedied by a single PCR or CE rerun. The assay’s overall first-pass QC success rate was 96.3%, and the second-pass QC pass rate was 99.8%. The most common reasons for QC failure were abnormal or missing ROX sizing peaks and low signals that prevented peak sizing or identification. Sizing ladder anomalies were typically corrected by rerunning the sample after reformulating the CE plate. Low signal anomalies were typically corrected by repeating the PCR reaction.
Comparator Assays
For in silico comparison of panel coverage, multiple targeted CFTR panels were used, including the Luminex xTAG CF39, CF60, CF71, and custom CF97 kits (Luminex Corp, Austin, Texas), TruSight Cystic Fibrosis 139-Variant Assay (Illumina Inc, San Diego, California), iPLEXPro CFTR Panel (Agena Bioscience Inc, San Diego, California), Elucigene CF-EU2 kit (Elucigene Diagnostics, Manchester, United Kingdom), eSensor CF kit (GenMark Diagnostics Inc, Carlsbad, California), the Devyser CFTR 68 kit (Devyser AB, Hägersten, Sweden), and the updated ACMG 2023 recommendation of 100 minimum variant set.11
For assay performance verification, multiple comparator methods were used, including Sanger sequencing, xTAG CF60, xTAG CF71, and CF-EU2 kits.
Screening Simulation
An in silico carrier screen simulation was used to evaluate panel performance. Variant allele frequencies from multiple ancestral groups within the gnomAD v2.1 (http://www.gnomad-sg.org/) data set were used. First, the observed gnomAD allele frequencies for all variants in each panel were extracted across the 7 populations defined in gnomAD 2.0.2. Next, 100 carriers were randomly simulated with the observed allele frequencies in the data set. A mock screen was performed with each respective panel to determine the number of carriers identified. This process was repeated 1000 times to generate 1000 simulations of 100 carriers each.
Samples
For the method comparison study, genomic DNA was isolated from 51 anonymized residual clinical samples of ethylene diamine tetraacetic acid–anticoagulated whole peripheral venous blood collected at 3 sites, 31 cell-line samples (Coriell Institute for Medical Research, Camden, New Jersey), and 2 contrived samples consisting of pooled synthetic DNA fragments. The whole blood and cell-line DNA together represented 51 of the kit-supported 63 variants plus PolyT/TG, and the contrived samples represented the remaining 14 assay-specific variants. No-template controls were included alongside all sample evaluations. Other studies used a subset of these samples, as appropriate.
This validation work leveraged cell lines from the Coriell Institute in multiple studies. These cell lines and DNA samples were obtained from the National Institute of General Medical Sciences Human Genetic Cell Repository, the National Human Genome Research Institute Sample Repository for Human Genetic Research, and the Coriell Institute for Medical Research at the US Centers for Disease Control and Prevention. A list of sample IDs used in these studies is included in Supplemental Table 4.
Primary Performance Analytics
The following variant-level accuracy metrics were evaluated: PPA, NPA, OPA, and PPV and zygosity agreement across SNVs and indels combined. Correct identification of PolyT/TG genotypes was also evaluated. Sample-level accuracy was defined as the percentage of correct genotype calls (ie, WT, 0 variants; heterozygous mutant [HET], 1 variant [carrier]; homozygous mutant [HOM Mut], 2 variants [affected]; or compound/multiple heterozygous mutant [COMP HET Mut], 2 or more variants [likely affected]) of all samples queried. For comparison purposes, genotypes were grouped as having 0, 1, or 2 or more variants, corresponding to WT, carrier, and affected genotypes. Sensitivity studies evaluated the assay’s ability to detect alleles with low CE peak heights (RFU). The effects of using different PCR and CE instrument configurations, operators, and DNA input quantities on assay output were evaluated.
Secondary Performance Analytics
Kit reagent and intermediate assay product stability assays were performed. Total assay time, comprising operator hands-on time and instrument run time, and pipetting requirements were measured. Secondary performance experimental details are provided in the Supplemental Materials text, and time-and-motion observations are presented in Supplemental Table 3.
Multisite Evaluation
The 3-site evaluation study included 129 unique clinical samples composed of genomic DNA isolated from whole blood (n = 125), uncultured amniotic cells (n = 1), cultured amniotic cells (n = 1), chorionic villi (n = 1), and fetal tissue (n = 1). Two sites were located in the United States, and 1 was located in Europe. Each site used unique instrumentation for PCR and CE; site A used a SimpliAmp thermal cycler with a 3730 Genetic Analyzer, site B used a Veriti thermal cycler with a 3500 Genetic Analyzer, and site C used a Veriti thermal cycler with a 3130 Genetic Analyzer. Each site compared results to their current comparator method (site A: Luminex xTAG CF60v2 kit; site B: Elucigene CF-EU2v1 kit; site C: Luminex xTAG CF60v2 kit with third tube for 97 variants total). Because the variant content of each assay panel is unique, CFTR variants detected by the CFTR PCR/CE assay that were not included in comparator panels were presumed to be WT, and variants detected by other panels but not included in the panel were excluded from analysis.
Statistical Analysis
Data were analyzed using JMP Statistical Software v15.0.0 (JMP Statistical Discovery LLC, Cary, North Carolina).
RESULTS
CFTR Variant Coverage by Different Targeted Panels
The CFTR PCR/CE panel targeted 65 pathogenic or likely pathogenic variants represented across diverse ethnic and ancestral groups. The panel variants account for the 92% prevalence catalogued in either the non-Hispanic White–predominant CFTR2 public database with more than 89 000 samples5 (updated April 2022) or a large 2019 genomic sequencing survey containing more than 115 000 patient samples with self-reported ethnicity. The latter more closely resembles the ethnically diverse US population (Figure 1, A and B).12
Pathogenic CFTR variant detection according to commercial kit and reference data sets. Three data sets with variant frequency and ethnicity or ancestry were used. A, Extrapolated ethnicity from the CFTR2 data set: ethnicity from 89 052 individuals within the CFTR2 public database5 collected across the United States, Canada, and Europe was extrapolated from reports by Sosnay et al17 and Bosch et al.26 B, Ethnicity breakdown: self-reported ethnicity was reported by Beauchamp et al12 across 72 847 US patient screenings. Patients for whom ethnicity was not reported or who had a family history of cystic fibrosis or infertility were excluded. C, Ancestry breakdown in gnomAD 2.1: ancestral populations reported by the gnomAD 2.1 data set16 across 141 456 whole-exome or genome sequences using population descriptions from gnomAD 2.0.2 that more closely matched those reported by Beauchamp et al.12
Pathogenic CFTR variant detection according to commercial kit and reference data sets. Three data sets with variant frequency and ethnicity or ancestry were used. A, Extrapolated ethnicity from the CFTR2 data set: ethnicity from 89 052 individuals within the CFTR2 public database5 collected across the United States, Canada, and Europe was extrapolated from reports by Sosnay et al17 and Bosch et al.26 B, Ethnicity breakdown: self-reported ethnicity was reported by Beauchamp et al12 across 72 847 US patient screenings. Patients for whom ethnicity was not reported or who had a family history of cystic fibrosis or infertility were excluded. C, Ancestry breakdown in gnomAD 2.1: ancestral populations reported by the gnomAD 2.1 data set16 across 141 456 whole-exome or genome sequences using population descriptions from gnomAD 2.0.2 that more closely matched those reported by Beauchamp et al.12
We compared panel coverage of 11 targeted CFTR panels or variant sets for both the CFTR2 data set and the large, ethnically diverse genomic data set from 2019 (Table 2). Coverage calculations for CFTR2 included 5T PolyT MAF for panels that detect PolyT (MAF, 0.363%) because this variant is pathogenic when it occurs in cis with R117H.10 The CFTR PCR/CE panel had the highest variant coverage based on the 2019 study data set, whereas the TruSight Cystic Fibrosis 139-Variant Assay had the highest coverage based on variant prevalence calculated from the CFTR2 data set. Although ethnicity across the entire CFTR2 data set is not available through the website, a 2013 study stated that 95% of the 31 727 cases where ethnicity could be identified were non-Hispanic White.17 This matches favorably with reporting from the CF Registry Annual Reports, where 93.4% of the 31 411 patients were listed as “Caucasian.” However, a recent New York State Department of Health report showed only 76% of those tested for CF were listed as “Caucasian,” suggesting that the testing population is more diverse than the non-Hispanic White–predominant population used to estimate frequencies and inform the design of many commercially available panels.18 This explains the reduced coverage of many panels when variant frequencies are estimated from large, diverse populations (Table 2).
A 2019 study that screened 381 000 individuals in the US population confirms these results.12 Of the top 5 CFTR variants catalogued as most frequent within each ethnic subgroup, the design of the CFTR PCR/CE panel studied here interrogates 4 substitution variants (R1070Q, I618T, V456A, and G970D) common in persons of African, East Asian, and/or Southeast Asian descent; these variants are not represented in panels of other commercial multiplex CFTR assays (Supplemental Table 1). Three variants were also prevalent in the 2019 study,12 showing that a small subset of critical variants is consistently responsible for pan-ethnic coverage gaps in many targeted CFTR panels.
CFTR Variant Coverage Determined by Simulated Carrier Screen
An in silico screening simulation was performed to compare carrier detection across different panels and ancestries using data available in the gnomAD 2.1 database. The gnomAD database contains more than 140 000 whole-exome and genome sequences with genetically determined ancestry (Figure 1, C).16 This resource expands the US-centric comparison of Beauchamp et al12 to a global frequency of likely pathogenic CFTR variants.
In 1000 simulated carrier screens across 100 carriers, each of the targeted panels correctly classified more than 85% of European (Finnish plus non-Finnish European) and more than 90% of Ashkenazi Jewish carriers based on median percent carriers identified (Figure 2). However, most comparator panels captured fewer than 85% of carriers from the Americas and Africa, and often as low as 20% to 30% of carriers from Asia. Indeed, commercially available targeted CFTR variant panels likely informed by CFTR2 data set frequencies are reported to have lower than expected coverage in non-White populations.19 In contrast, the CFTR PCR/CE panel content covered 95% of carriers overall across an aggregate of all populations in the data set, largely because of improved performance among carriers from Africa and Asia. This shows that the variants chosen for this panel can help address population disparities in carrier screening for CFTR variants, and further, that the number of variants in a panel is not a good indicator of coverage, especially in diverse populations.
Simulated carrier screens across different targeted CFTR panels and populations using gnomAD allele frequency of (likely) pathogenic variants. Carriers (n = 100) were simulated across variants in all panels, with CFTR variant frequency identified within 141 456 whole-exome or genome sequences in gnomAD 2.1.16 For each panel, carrier detection was determined based on variants in the specific panel. The process was repeated 1000 times, and results were compared across both panel and population. The gnomAD population naming conventions are used, including Global (all gnomAD), Ashkenazi Jewish (ASJ), Latino/Admixed American (AMR), African/African American (AF), East Asian (EAS), South Asian (SAS), and Finnish plus non-Finnish European (FIN + NFE). Assay panels are identical to those in Table 2.
Simulated carrier screens across different targeted CFTR panels and populations using gnomAD allele frequency of (likely) pathogenic variants. Carriers (n = 100) were simulated across variants in all panels, with CFTR variant frequency identified within 141 456 whole-exome or genome sequences in gnomAD 2.1.16 For each panel, carrier detection was determined based on variants in the specific panel. The process was repeated 1000 times, and results were compared across both panel and population. The gnomAD population naming conventions are used, including Global (all gnomAD), Ashkenazi Jewish (ASJ), Latino/Admixed American (AMR), African/African American (AF), East Asian (EAS), South Asian (SAS), and Finnish plus non-Finnish European (FIN + NFE). Assay panels are identical to those in Table 2.
Sample-Level Accuracy
We assessed the accuracy of the CFTR PCR/CE kit at the sample and variant levels. Sample-level accuracy was defined as the percentage of correct genotype calls (ie, WT, HET, HOM Mut, or COMP HET Mut) from 489 known-identity samples (not shown). The assay correctly identified each of the 3 genotype subgroups with positive agreement of 95% or greater, yielding an overall agreement of 97.1%. Of a total of 12 incorrect sample genotypes, 11 (91.7%) were due to incorrect zygosity calls; all variants were correctly reported as positive or negative, but they were reported as homozygous instead of heterozygous or vice versa. Of these 11 calls, 6 were from a single sample and PCR analyzed on 6 CE configurations where G85E was called homozygous instead of heterozygous because of a large exon deletion (CFTRdele2,3) in this region on the opposite chromosome that removed the WT G85E allele. Thus, the presence of the CFTRdele2,3 deletion caused G85E to be homozygous because there was only 1 copy of the allele that contained the variant. An additional 4 of 11 incorrect calls were also from a single sample and PCR analyzed on 4 of 6 configurations; in these cases, D1152H was called homozygous instead of heterozygous because of missed detection of a WT peak with unknown root cause. Further refinements to the software algorithm may improve these results.
Variant-Level Accuracy
Using the same data described above under sample-level accuracy to assess variant-level performance, more than 30 000 variant calls were determined using multiple CE platforms (Table 3). Accuracy in variant identification using the assay was consistently high across various CE instruments, showing an overall PPA of 99.7% and a PPV of 99.9%, with no more than a single miss for any particular SNV or indel. The cumulative NPA and OPA values were both 100%. Zygosity agreement was 100% across CE platforms, and PolyT/TG agreement with expected results was 99.4% overall and 100% on 3 of 4 CE platforms tested. Thus, the pre-established acceptance criterion of achieving 95% or greater accuracy was met for all variant-level evaluations (overall PPA, PPV, NPA, OPA, and percent agreement for zygosity and PolyT/TG presence).
In related studies, the assay accuracy was compared when DNA samples were run on 4 commonly used thermal cyclers (models 9700, ProFlex, and Veriti from ThermoFisher; and model C1000 from Bio-Rad Laboratories). A total of 6136 sample-replicate determinations were performed on the 9700 instrument, and 3068 sample-replicate determinations were performed on each of the remaining 3 instruments. Amplified products were analyzed on a 3500xL CE system. Sample DNA was derived from 6 clinical samples and 6 cell line samples, and DNA inputs were either 20 or 60 ng (reflecting kit-specified range). Across all 4 instruments, the PPAs and NPAs were uniformly 100%, with no erroneous calls. Taken together, this demonstrates assay performance across 24 different instrument configurations.
Multisite Evaluation
To evaluate the performance of the assay in different clinical laboratory settings, 3 sites with unique instrumentation configurations and isolation methods each tested a unique cohort of samples and compared results from the assay to their current comparator method (see Materials and Methods). The study included a total of 129 unique clinical sample measurements representing 39 unique variants detected by the kit, or approximately 85% of the total 92% variant coverage provided by the assay.12
Of 129 unique clinical samples tested at 3 external sites, 128 of 129 (99.2%) produced correct sample-level genotype calls (Table 4). The single incorrect genotype was due to an incorrect zygosity, where an expected heterozygous 2184delA call was called homozygous by the assay. Overall, these data indicate that assay performance is robust, generating consistent results across different laboratory environments with different thermal cyclers, CE instruments, isolation methods, and comparator methods.
Sensitivity
The sensitivity study assessed the CFTR PCR/CE assay’s ability to detect alleles with relatively lower peak signals when input DNA was at the low end of the kit’s 20- to 60-ng range. A total of 11 lower-peak-height alleles were selected for this study: I507del-mut, R1158X-mut, R75X-mut, E60X-mut, Q493X-mut, V456A-mut, Q493X-wt, R334W-wt, L206W-wt, 1677delTA-wt, and V520F-wt. The study was performed by 1 operator with each sample run in quadruplicate on 2 instrument configurations: Veriti thermal cycler + 3500xL (50-cm) CE and 9700 thermal cycler + 3130xl (36-cm) CE. Samples included previously characterized DNA from 7 Coriell cell lines and 1 whole blood specimen, using 20 ng of DNA per PCR. All 11 alleles with lower peak heights were successfully detected by the assay software using both instrument configurations (Figure 3; Table 5).
Detection of CFTR variants with low peak heights in limited-quantity DNA samples. Samples included previously characterized DNA from 7 cell lines and 1 whole blood specimen, using 20 ng of DNA per polymerase chain reaction (PCR). All of the 11 CFTR variants in the CFTR PCR/capillary electrophoresis (CE) kit with low-peak heights expected on CE were successfully identified in DNA samples at the low end of the kit’s DNA input range (ie, 20 ng of DNA), using 2 instrument configurations: 9700 cycler + 3130xl (36 cm) CE array and Veriti cycler + 3500xL (50 cm) CE array. Relative fluorescence units (RFUs) were detected by both instrument configurations. The 3500xL RFUs were ≈3-fold larger than the 3130xl RFUs because of the different dynamic ranges of each instrument’s laser. Abbreviations: MUT, mutant peak; WT, wild-type peak.
Detection of CFTR variants with low peak heights in limited-quantity DNA samples. Samples included previously characterized DNA from 7 cell lines and 1 whole blood specimen, using 20 ng of DNA per polymerase chain reaction (PCR). All of the 11 CFTR variants in the CFTR PCR/capillary electrophoresis (CE) kit with low-peak heights expected on CE were successfully identified in DNA samples at the low end of the kit’s DNA input range (ie, 20 ng of DNA), using 2 instrument configurations: 9700 cycler + 3130xl (36 cm) CE array and Veriti cycler + 3500xL (50 cm) CE array. Relative fluorescence units (RFUs) were detected by both instrument configurations. The 3500xL RFUs were ≈3-fold larger than the 3130xl RFUs because of the different dynamic ranges of each instrument’s laser. Abbreviations: MUT, mutant peak; WT, wild-type peak.
Specificity
For the analytic specificity analysis, a subset of data generated on the Veriti thermal cycler and resolved on a 3500xL CE platform (50-cm capillary length) was pooled from the within-laboratory precision, DNA input, thermal cycler equivalency, and method comparison studies.
To assess analytic specificity (exclusivity and inclusivity) for all variants detected by the assay, percent agreement with Sanger sequencing was determined with all homozygous WT variant-level zygosity measurements for exclusivity (n = 963) and with all homozygous variant-level zygosity measurements for inclusivity (n = 113). Across all variants, 100% of zygosity measurements agreed with Sanger sequencing (not shown). These data demonstrate perfect specificity for all variants within the assay.
Within-Laboratory Precision
A single-site precision study evaluated DNA from 6 clinical samples and 5 cell lines, assayed in duplicate by 2 operators on 6 different days (n = 24 observations per sample), using Veriti thermal cyclers and 3500xL CE analyzers. Of 16 334 total variant calls, the PPV, PPA, NPA, OPA, and zygosity agreement were all 100% for each operator, suggesting high interoperator precision. Additionally, the overall genotype-level agreement was 100% (264 of 264).
Additional Studies
The effect of DNA input quantity, stability (reagents and amplified products), and time-and-motion analyses were also performed and are detailed in the Supplemental Materials. Briefly, assay performance at or above specification was achieved with a variety of input DNA quantities (15–70 ng per reaction), kit reagents, and PCR products, and remained stable in various time- and temperature-dependent storage scenarios. Total assay time averaged 4 hours 33 minutes, comprising 43 minutes of operator hands-on time and 230 minutes of hands-off instrument time.
DISCUSSION
The risk of CF and the relative frequency of pathogenic CFTR variants differ across ethnicities and geographies.10–15 This presents a challenge for designing targeted panels that detect a high percentage of carriers, or even CF patients, in diverse populations. Originally, the ACMG recommended that panels include variants with a pan-ethnic frequency of 0.1% or greater within the general US population.7 A 2004 report highlighted a clear disparity in clinical sensitivity (detection of a fetus with 2 identifiable variants) when using assays that only interrogate the CF23 variants; cumulative MAF coverage varied from a high of 89% for Ashkenazi Jewish couples to a low of 24% for Asian-American couples.3 That same year, the ACMG reissued its recommendation supporting use of the abbreviated CF23 panel.9 However, our understanding of CFTR variants and their frequencies within the diverse US population20 has changed drastically with the widespread application of NGS testing, reflected by recent investigations that have provided critical updates to the prevailing pathogenic CFTR variant profile within the United States.12,13
Whereas the ACMG recently expanded its minimal carrier screening panel recommendation to include 100 variants designed to detect 95% of CF carriers across 6 ancestral groups,11,16 it excluded all variants of varying clinical consequence (VVCCs) as defined by CFTR2.5 A small number of these VVCCs have recently been shown to be highly prevalent in non-White populations and are broadly considered pathogenic, including in CFTR2, where they are associated with moderate to severe CF-related phenotypes.5,12,13,19 Because of this, the updated ACMG minimum set of 100 variants (ACMG 202311 ) represents less than 87% of the carriers observed in 1 recent population-level study in the United States published in 2019 (Table 2).12
The importance of accurately screening for ethnicity-specific CFTR variants is highlighted by the 2019 NGS study,12 which describes 115 571 individuals who were screened in the United States during 2017–2018. In that study, the authors indicated that testing only for the CF23 would have missed identifying 31% of at-risk pregnancies.12 Notably, in persons who self-identified as Hispanic, for whom the risk of CF is relatively high, screening limited to the CF23 would have missed 75% of affected pregnancies.12 Whereas the updated ACMG 2023 minimum set of 100 variants represents a significant improvement toward the goal of more equitable coverage by ancestry, the exclusion of highly prevalent variants solely based on CFTR2 status nevertheless resulted in less than 90% detection of carriers based on allele frequency estimates from this study.12
A 2020 NGS survey of 381 014 individuals in the United States identified 481 known or suspected pathogenic CFTR variants beyond the standard CF23 and noted that 44% of carriers would have been missed by CF23 screening alone.13 That study also illuminated the large disparity in frequencies of the top 5 pathogenic CFTR variants across different ethnic groups, where 4 of the 5 most common CFTR variants detected in individuals identifying as East Asian and Southeast Asian would not have been detected using the CF23 panel. Among African Americans, 3 of the 5 most common CFTR variants they detected would not have been captured by CF23 screening. Despite recent updates, the ACMG 2023 minimum set of 100 variants still excludes 4 of 5 top variants in East Asians, 3 of 5 top variants in Southeast Asians, and 3 of 5 top variants in African Americans in that study.13 These observations provided special impetus to revisit whether routine CFTR screening should include additional targeted variants that are at increased incidence and clinical relevance in non-White ethnic groups, especially for variants considered VVCC by CFTR2 with evidence of pathogenicity.5,10,12–15
Contemporary medicine strives to achieve equitable health care outcomes for all patients regardless of ethnic, racial, and socioeconomic identity.21 Cystic fibrosis newborn screening has been uniformly offered across the United States since 2010; however, targeted variant identification in non-White populations still remains inequitable.22 Screening models in many regions, including the United States, Canada, and Europe, have historically been based on non-Hispanic White–centric CFTR variant profiles that do not accurately reflect the ethnic composition of their populations and would benefit from reassessment and revision.22–24 Advances in CFTR variant identification and multiplex analytic methodologies have reduced testing costs25 and can provide more equitable comprehensive screening against CFTR variant panels that have been expanded to better reflect demographics and associated variant frequencies.10,14,15
Whereas comprehensive NGS sequencing can circumvent the challenges of diverse CFTR variant profiles by detecting all variants,12,13 it is both laborious and costly. Additionally, NGS approaches are still limited by clinical annotation of variants and a shifting landscape of variants of unknown significance. The use of NGS for carrier screening may become necessary as laboratories begin to accommodate recent ACMG guideline updates recommending a minimum set of 100 variants11 ; however, the practical limitations mentioned above may be a barrier to this progress. In contrast, targeted assays can provide streamlined workflows with straightforward analysis, but their content must represent the intended test population to be effective. To maximize the clinical utility of updated pathogenic CFTR variant databases that more accurately reflect variant distributions across ethnicities, targeted screening and diagnostic assays need to be carefully designed to capture those variants.
The CFTR PCR/CE panel described herein accurately and reliably recognized variants that represent 92% of pathogenic or likely pathogenic variants circulating in the contemporary pan-ethnic US population based on recent large-scale sequencing studies.12,13 Percent coverage was better than all 10 targeted assays that were compared, resulting in 4.0% to 13.2% higher coverage for the CFTR PCR/CE targeted content. A total of the 3 the 10 comparator assays targeted a larger number (n = 74 to 139) of variants, but in each case this did not translate into better coverage for the current US CFTR population profile.12 The panel also represents 92% of variants in the predominantly non-Hispanic White CFTR2 public database that less optimally reflects current US ethnic demographics.5 In simulation studies, the panel covered 95% of carriers overall across the global population, resulting in 6.95 to 19.0% higher carrier detection rate than all other comparator assays.
By contrast, all 10 comparator assays attained better cumulative coverage when queried against the CFTR2 data set than they did against the more ethnically inclusive database. This result suggests a concerning miss rate when assessing CFTR variants in non-White within the United States and other geographic regions. Whereas the prevalence of CF in some ancestries, such as Asians, is much lower than in non-Finnish European/Ashkenazi Jewish populations, providing equitable coverage nevertheless requires maintaining similar detection rates across populations regardless of prevalence. Indeed, improved equity was the primary objective of the latest ACMG carrier screening recommendations, and is an approach recognized by medical organizations responsible for CF and other genetic disease testing guidelines.11 Thus, inclusive CFTR variant detection in the United States relies on accurately targeting the variants that are predominant throughout the intended test population, especially when it comprises several different genetic ancestries.
The assay exceeded all pre-established laboratory performance criteria and consistently generated reliable high-fidelity data for the targeted 65 CFTR variants. At the variant level, accuracy metrics, including PPV, PPA, NPA, and OPA, were greater than 95% across all SNVs and indels in all studies. Similarly, zygosity agreement and PolyT/TG agreement all exceeded 95%. Consistently, correct genotype calls reflected sample-level accuracy.
Whereas our study shows the assay is robust across conditions tested here, the number of PCR cycles had to be optimized to accommodate all thermal cyclers tested. Furthermore, the algorithm developed for analysis relies on machine learning, and the rarity of some CFTR variants meant that the algorithm had to be trained on synthetic samples for some of the low-frequency variants. Thus, additional optimization would be necessary to accommodate instrument platforms not tested here.
Assay fidelity was maintained across 24 different instrument configurations and when performed by different operators across different sites. Variability due to different instrument configurations, workflows, and operators using a common sample panel was low. The assay accommodated varied DNA input amounts commonly observed from standard isolation methods, and the kit reagents and amplicon intermediary products were tolerant of diverse storage conditions, including repeated temperature cycling. The assay has a streamlined workflow with dedicated analysis software and can generate results in as few as 5 hours.
CONCLUSIONS
The panel provides improved coverage with 65 disease-causing CFTR variants, including variants that are prevalent in minority subgroups of the contemporary United States and global populace. The assay exceeds acceptance criteria for diverse analytic performance metrics, is compatible with multiple commonly employed PCR and CE platforms, and can provide CFTR genotyping results within 5 hours. Routine use of targeted CFTR variant assays that integrate variant frequency data across diverse ancestries has the potential to improve the detection of both CF carriers and patients and help address testing inequities.
The authors are grateful to Sarah Edelmon, Pranesh Rao, Shobha Gokul, Liangjing Chen, Jon Kemppainen, and Adrian Lara (Asuragen) for assay development and verification testing efforts. We also acknowledge Jessica Larson, Brian Haynes, and Ted Markulin (Asuragen) for assistance in modeling the detection of carriers across targeted CFTR panels. The authors thank Matt Silverman, PhD (Biomedical Publishing Solutions, Panama City, Florida), for scientific, analytic, and writing assistance.